AI Pulse · Updated every Monday

The weekly AI briefing for working professionals.

Five minutes. Zero hype. Every Monday I hand you the AI news that actually matters to your career — and the one thing to do about it this week.

Skip to this week →
This week Week of April 14–20, 2026

What actually happened in AI this week.

Curated from the noise. Every story ends with why it matters for your career specifically.

★ Headline story

Claude Opus 4.7 launches April 16 — Anthropic reclaims the top LLM spot

Anthropic released Claude Opus 4.7 on April 16, scoring 64.3% on SWE-bench Pro. Key additions: agentic task budgets for long-running workflows, a /ultrareview command for senior-level code review, and Claude Design — a Labs tool for creating slides, prototypes, and one-pagers visually. Available everywhere (claude.ai, API, Bedrock, Vertex, Microsoft Foundry) with pricing unchanged.

Why this matters to you: Opus 4.7 is built for the multi-step tasks your company is about to delegate to AI. Understanding what it can — and cannot — do puts you ahead of colleagues who will just react to the rollout.
Models

GPT-5.4 vs Gemini 3.1 Pro: two models fighting for the benchmark crown

GPT-5.4 leads the Artificial Analysis Intelligence Index on knowledge work (83% GDPval) and coding. Gemini 3.1 Pro leads on abstract reasoning and science, and ships with a 2-million-token context window — enough to process a full year of meeting transcripts or a 400-page contract in one prompt.

Why this matters to you: Context windows at this scale change what "research" means for a professional. Paste a full briefing document and ask for gaps — it's faster than a junior analyst. Try it this week.
Jobs market

Oracle axes ~30,000 jobs in April — total Q1 tech cuts exceed 85,000

Oracle laid off roughly 30,000 employees (18% of global headcount) in late March and early April to fund its $300B OpenAI data-centre contract. Combined with earlier Q1 cuts, the tech sector has shed over 85,000 jobs since January — nearly half explicitly attributed to AI and workflow automation.

Why this matters to you: Companies are trading headcount for AI infrastructure. The safest position is being the professional who works alongside those systems — not the one they replace.
Productivity

Stanford's 2026 AI Index: AI capability is accelerating, not plateauing

Stanford's annual report confirmed that AI progress on key benchmarks accelerated in 2025–2026, reasoning model performance doubled year-over-year, and the cost of running a leading-edge model dropped 280× over three years. Public concern about AI's impact on jobs hit its highest recorded level.

Why this matters to you: The speed of change is real and verified. Treating AI fluency as a weekly discipline — not an annual training — is now the rational professional response.
Enterprise

White House releases National AI Policy Framework — federal rules land this month

The Framework (effective April 2026) calls for AI governance through existing agencies rather than a new regulator, federal preemption of conflicting state laws, and privacy protections for children. The RAISE Act also requires frontier AI developers to publish safety and transparency data.

Why this matters to you: Your company's legal, compliance, and HR teams are acting on this now. Understanding AI governance at work — acceptable use, data handling, disclosure — makes you the person invited into those conversations.
Creative work

Adobe Firefly AI Assistant orchestrates all Creative Cloud apps from one prompt

Unveiled at Adobe Summit 2026, the Firefly AI Assistant lets users describe an outcome in plain language and have the agent execute multi-step workflows across Photoshop, Premiere, Lightroom, Illustrator, and Express — no menu hunting required.

Why this matters to you: Marketing and content teams: your "how fast can you ship a campaign" benchmark just reset. The professionals who master prompt-to-asset workflows will outpace the ones still hunting menus.
🛠 Tool pick of the week

Claude Design — make slides, prototypes, and one-pagers with AI in minutes

Launched alongside Opus 4.7, Claude Design (labs.anthropic.com) lets non-designers describe a visual output in plain language and get a polished prototype back. Paste your meeting agenda or project brief, say "turn this into a one-page visual summary I can share before the call," and it's done in under 5 minutes. Editable and exportable. No design skills required.

Why this matters to you: If you've ever spent an hour making a deck look presentable, this is the tool to try first this week. The entry bar is just: can you describe what you want?
💡 Career tip of the week

Post one "before AI vs. after AI" workflow comparison on LinkedIn this week

Pick one real task from your past week — a report, an analysis, a deck, an email thread. Write a two-paragraph post: paragraph one is how you used to do it (time, steps, friction); paragraph two is how AI changed it. Include one specific number: minutes saved, pages processed, emails cut. Takes 15 minutes.

Why this matters to you: In an "AI-washing" layoff market, visible proof of AI fluency is the single highest-leverage career move you can make in under 15 minutes. Hiring managers aren't impressed by "I use AI" — they want receipts.

Get next Monday's Pulse in your inbox.

One email a week. The free AI toolkit on signup. Unsubscribe any time.

Past briefings.

Every Monday since launch. The archive compounds — come back weekly to see how the story evolves.

Earlier editions
Week of April 13, 2026

Anthropic's revenue passes OpenAI — $30B vs $25B annualized

Anthropic announced its revenue run rate crossed $30B, overtaking OpenAI for the first time, with over 1,000 enterprise customers spending >$1M/year on Claude. Both Claude Opus 4.7 and Gemini 3.1 Ultra were announced this week.

What we covered: Enterprise AI spending race, Opus 4.7 & Gemini 3.1 Ultra launches, Q1 layoff data, Google Sheets AI productivity gains, Adobe Firefly launch.
Week of April 6, 2026

OpenAI, Anthropic, Google unite against model copying

The three US AI leaders announced a coordinated intelligence-sharing effort via the Frontier Model Forum to stop Chinese firms (DeepSeek, Moonshot, MiniMax) from adversarial distillation of their models.

What we covered: Geopolitics of AI, MCP crossing 97M installs, Anthropic's Coefficient Bio acquisition, life sciences expansion.
Week of March 30, 2026

GPT-5.4 Thinking + Gemini 3.1 Pro benchmark sweep

OpenAI's GPT-5.4 took the versatility crown; Claude Opus 4.6 and Gemini 3.1 Pro both topped 50% on Humanity's Last Exam. Frontier model progress visibly accelerated.

What we covered: What "reasoning models" actually do in practice, benchmark literacy, and when to pay for Pro vs use free tiers.
Week of March 23, 2026

NVIDIA GTC: agentic AI goes production

Fortune 500 companies announced production agentic deployments across manufacturing, logistics, and finance. The conversation shifted from benchmarks to business value.

What we covered: What an "agent" really is (in plain English), the five agent patterns showing up at work, how to evaluate them.

More editions coming as the archive grows. Subscribers get every new one delivered.

Never miss a Pulse.

Subscribe once. I'll send the weekly briefing every Monday — plus the free AI toolkit today.